Ecological theory suggests that early adolescent peer groups regularly form which encourage problem behavior, leading to even closer, deviance-prone peer groups, and so on in a self-reinforcing cycle. We call this process toxic peer contagion. It is best conceptualized as a Nonlinear Dynamic System (NDS). We propose to develop NDS models that explain the onset to and growth of problem behaviors (antisocial behavior, aggression, substance use) in middle school peer social ecologies, and particularly to examine toxic peer contagion as an explanatory hypothesis. The study will be conducted in 21 sixth- to eighth-grade middle schools involved in an already underway, NIDA- funded group-randomized trial of the Positive Behavior Support (PBS) program. A major goal is to investigate variation in toxic peer contagion differences between schools, especially between PBS program and control schools. The project requires collecing social relationship (social network) and relevant behavioral data on multiple occasions, and observing the pattern of interdependence between them over time. Assessments will be conducted four times per school year, for two consecutive entering sixth-grade cohorts, through the end of their eighth-grade year in each of the 21 participating schools; thus, 12 observations per cohort. School-level models will be created using SIENA, a recently developed statistical methodology specifically designed to estimate models of evolving social networks and co-evolving behavior. Also, two-level models will be developed when effects vary significantly across schools. In order to accomplish these ambitious goals with this large, diverse sample of schools and students, it is necessary to perform several major upgrades to the SIENA computer software, improving its estimation performance by porting it to multiprocessor computing environments, and enhancing its ability to include interaction effects, full- information multilevel models, and cross-level interactions. For this purpose, we are partnering with the researchers who developed SIENA to make the upgrades necessary for project goals. The results of this project will include: (a) estimates of the prevalence of toxic peer contagion in a varigated sample of middle schools; (b) conditions under which it is most and least likely to occur; (c) the effectiveness of PBS as a way to discourage it; and (d) recommendations for peer-directed interventions for dealing with this problem. This project will examine the effect of a group-randomized trial of an already funded school-level positive behavior intervention on the over-time interdependence between early adolescent problem behavior and peer group dynamics. This process, developed from theories of social ecologies and development, is toxic peer contagion. Nonlinear dynamic models of affiliation-behavior interdependence will be statistically estimated from longitudinal (4x/yr, 6th to 8th grade) data using SIENA, a computer program specifically developed to model such processes. Results will explain the role of peer affiliations in development of problem behavior, the effect of positive behavior support on this linkage, and suggest new avenues for future interventions.